e t o u q Report of the HEI Diesel r o Epidemiology Panel - - PowerPoint PPT Presentation
e t o u q Report of the HEI Diesel r o Epidemiology Panel (Part II): e Diesel Epidemiology and t i c Lung Cancer t o n Introduction and Framing o Katherine Walker, Health Effects Institute D e Outline t o u q A short
2
associate exposures to older technology diesel engine exhaust with increased rates of lung cancer
guidelines limited by exposure assessments
use of the then available epidemiologic studies in railroad workers and in teamsters for quantitative risk assessment
3
Better measures of exposure
susceptibility
Better models of exposure
exposure is likely to occur, and current and historical data regarding emission sources.
exposures ……
Better study designs for exposure-response
respect to magnitude, frequency, and duration, rather than solely by duration of employment.
range of exposures to provide a base for understanding the relation between exposure and health effects.
quantified where possible;
power calculations and exposure–response analyses.
this disease.
with smoking histories will strengthen the interpretation of results.
Research Needs for Quantitative Risk Assessment (HEI 1999, 2002 )
4
KEY COMPONENTS:
CAVEATS:
(ACES)
less developed countries)
5
National Cancer Institute/ National Institute of Occupational Safety and Health (NCI/NIOSH)
6
DEMS Truckers Design Cohort and Nested Case- Control Cohort Questionnaire Yes, individual level risk factors No Population 8 U.S. non-metal mines (limestone, trona, salt, potash) 12,315 miners: 96% male, 88% white 139 U.S. Trucking terminals 31,135 worker: 100% male, 85% white Lung cancer 198 lung cancer cases, 563 controls matched on mine, sex, race, and birth year 779 lung cancer cases End of follow-up 1997 2000 Metric of personal exposure Respirable elemental carbon (REC) ≤3.5 µg/m3 Submicron Elemental Carbon (SEC) ≤ 1 µg/m3 7
8
exposure
for duration of work – healthy worker effect
Entire Cohort
Hazard Ratio=1
Excluding Mechanics 9
10
Odd’s Ratio=1
Silverman et al. 2012 All Subjects
Charge Questions Appoint Panel Evaluation
analytical data sets
Draft report Peer Review Final Report
2015
Public workshop
We are here 11
1. Reviewing the findings of the 1999 HEI Special Report on epidemiology and risk assessment 2. For recent epidemiologic studies, reviewing their design, data, and exposure estimates, … analyzing such data as needed. 3. Exploring whether the data from these new studies enables analyses to extend concentration–response relationships to lower ambient concentrations 4. Identifying data gaps and sources of uncertainty. 5. Making recommendations about extension or further analyses of existing data sets. 6. Making recommendations, if necessary, about the design of new studies that would provide a stronger basis for risk assessment. 12
Daniel Krewski, PhD, Chair
Professor and Director of the R. Samuel McLaughlin Centre for Population Health Risk Assessment at the University of Ottawa
Paul Demers, PhD
Director, Occupational Cancer Research Centre, Cancer Care Ontario and Professor, Dalla Lana School of Public Health, University of Toronto
David Foster, PhD
Professor Emeritus, Department of Mechanical Engineering, University of Wisconsin Madison
Joel Kaufman, MD, MPH
Professor, Environmental and Occupational Health Sciences, Medicine and Epidemiology; School of Public Health and School of Medicine, University of Washington
Jonathan Levy, ScD
Professor and Associate Chair, Department of Environmental Health, Boston University School of Public Health
Charles Poole, ScD, MPH
Associate Professor, Department of Epidemiology, University
Nancy Reid, PhD
University Professor of Statistics, Canada Research Chair in Statistical Theory and Applications, University of Toronto
Martie van Tongeren, PhD
Director, Centre for Human Exposure Science, Institute of Occupational Medicine, Edinburgh, Scotland, UK
Susan R. Woskie, PhD, CIH
Professor, Department of Work Environment, University of Massachusetts-Lowell.
13
responses
Rodricks/Environ Corp.
14
Analytical data sets
NCI and NIOSH, with IRB approval
Ottawa
cannot be linked
downloadable on-line from NIOSH website
Additional opportunities
15
“Reanalysis of the DEMS Nested-case
control study of lung cancer and diesel exhaust: suitability for quantitative risk analysis”
Crump et al. (2015) *
exposure assignments
main CPH models
(WLMs)
underground workers
individual mines
“Diesel engine exhaust and lung cancer mortality ---time-related factors in exposure and risk” Moolgavkar et al. (2015)*
modeling with TSCE model
underground workers
analysis of risk using TSCE and CPH models
16
* Funded by consortium of companies led by the Engine Manufacturers Association
Research-Based Risk Assessment Risk Management Data Streams
Human
Animal Mechanistic Pharmacokinetic
distribution, metabolism, excretion
Exposure measurements, predictions, biomonitoring Hazard Identification Exposure Assessment Exposure-Response Assessment Characterization of Risk and Uncertainty Regulatory
Evaluate consequences of
Agency decisions and actions
Stakeholder input 17
IARC, others
Risk characterization: modeling, assumptions, adjustments uncertainties:
demographics
personal exposures over a lifetime
levels and composition Image of general population What is the observed risk of lung cancer? What is the predicted risk of lung cancer? Other relevant data methods, and analyses 18
Epidemiologic evaluation
strengths/limitations of the study in relation to the hypotheses which it was designed to test? Risk assessment
the prediction of risks associated with different levels of exposures, in different populations?
19
hypotheses, including adequate power and precision, the appropriate study population, and plans for evaluation of effect modification and control for confounding variables;
complete reporting of results;
assignment of exposure;
provides some insight to the magnitude and potential influence of key uncertainties in exposure assignment, and that is blind to identification of health outcomes;
a range of plausible alternatives, including biological relevance; and
assumptions in the design and analysis of the study.
20
and Other Factors (Jonathan Levy, Boston University School of Public Health)
Exhaust (Paul Demers, Occupational Cancer Research Center, Canada)
University of Ottawa)
21
Understanding the Potential Influence of Smoking, Radon, and Other Factors
Jonathan Levy Boston University School of Public Health
2
distorts the observed association between an exposure and disease
disease
correlations among exposures or distributions in comparison groups
exposure and a disease varies by levels of a third factor
controlled (by design or analysis); effect-measure modification cannot be controlled but should be understood
3
response functions by misstating the true association
biased concentration-response functions if effect modifiers are not explored and are distributed differently in the target population than in the study population
characterized but do not preclude use of the studies in quantitative risk assessment
4
5
individual smoking status
survey to stratified random sample of 11,986 current or recently retired employees of three companies
current/former/never smokers, the authors constructed adjustment factors by job title, which ranged from 0.92 to 1.17
influence on findings given small adjustment factors and socioeconomic similarity of cohort
Truckers is a limitation but not one that should preclude use of the study in quantitative risk assessments
6
previous findings that duration of employment was associated with reduced lung cancer mortality, posited to be a sign of survivor bias (control for a negative confounder)
created potential interpretability challenges given role of duration of employment in cumulative EC metrics
exposure and outcome
ventilation, job location in the terminal, and background exposures (predicted by local weather, proximity to major road, land use, and region) 7
pollutants that also have associations with lung cancer (e.g. silica, asbestos, radon, respirable dust, non-diesel PAHs)
8
subjects and by work location, by quartile of REC exposure, and by tertile of cumulative REC
monotonic, with diminished effect in ever-underground workers
smoking status, intensity, and location
was incorporated in the models
location and smoking that might be more informative for risk assessment
9
Silverman et al. 2012 Silverman et al. 2014 HEI Panel Exposure metric: Smoking status: Interactions: Average REC, lag 0, 15 yr Cumulative REC, lag 0, 15 yr Duration of REC exposure (yrs) Never, former, current, unknown; Intensity None (Smoking status and work location were combined in the analysis) Average REC, lag 15 yr Cumulative REC, lag 15 yr Status–Duration; Status–Pack-years; Status–Packs/day and duration None (Smoking status and work location were combined in the analysis) Average REC, lag 0, 15 yr Cumulative REC, lag 0, 15 yr Status–Duration; Status–Packs; Status–Pack-years; Status (Never, former, current, unknown) and Duration (continuous); Status and Packs/day (continuous); Status and Pack- years (continuous) Location of employment (ever underground/surface
packs/day, and pack- years as continuous variables
[1]
10
11
concluded they changed point estimates of ORs by ≤ 10% so did not include in final models.
given work duration in the model.
these levels have estimated effects substantial enough to create potential for appreciable confounding, and exposures correlate with underground status (no radon exposure at surface)
given significant correlation between cumulative REC and cumulative radon
12
Facility Mine Type % values <LOD Mean Area Concentration (pCi/L) Mean Area Ever-UG workers (Working Level)* A Limestone 15% 3.5 0.009 B Potash 56% 3.0 0.017 D Potash 61% 3.3 0.016 E Salt 30% 4.5 0.016 G Trona 76% 4.2 0.017 H Trona 85% 2.1 0.008 I Trona 80% 2.8 0.008 J Potash 62% 2.2 0.009 *Working level = a measure of exposure to radon and daughters. Source: Attfield et al. 2012. USEPA Residential action level = 4 pCi/L
13
cancer, the Panel concluded that there was no evidence indicating substantial confounding that would invalidate the application of the Truckers or DEMS studies for quantitative risk assessment
would be the ability to understand the implications of any information gaps or differences in risk factors (i.e., smoking status and intensity) on the degree of uncertainty in the exposure-response relationship for the target population
14
2
3
4
5
1960 1970 1980 1990 2000
Post-1971 1988 – 1989: EC data (Zaebst et al., 1991) 1971 – 2000: Monthly New Jersey COH data Pre-1971 2001 – 2006: Study surveya
a ~ 4000 personal/area samples (8-12 hrs) for EC in PM1.0
Measurements in 36 of 139 large terminals (and 44 nearby small terminals)
1971 – 2000: Work histories 6
7
.5 1 1.5 2 2.5 Multipliers 1985 1990 1995 2000 year CA COH NJ COH PM NJ PM CA PM US
Comparison of Background Multipliers
Ratio of Zaebst et al (1991) median background to current study (2.2). 8
9
10
1950 1960 1970 1980 1990 2000
1994: Feasibility study (Stanevich et al., 1997)
1950 – 1998: Work histories 1960 – 1998: Diesel equipment inventories 1947 – 1967: Dieselization 1970 – 1998: Mine ventilation maps
1975 – 1998: MIDAS mine inspection records (stationary CO, other gases)
1998 – 2001: DEMS Survey
(dust, EC, OC, gases) at 7/8 mines 1976: MESA Air Monitoring Survey
Validation datasets Determinants
1990’s: New engine technology
REC data Diesel trends 11
12
13
specific intercepts & slopes
15
16
Crump et al. (2015), Figure 1
*loader operator for mine A
17
18
19
20
Better measures of exposure
√ √
the complex mix of diesel exhaust emissions.
√ √
susceptibility are needed.
X X
Better models of exposure
area monitors placed where diesel exposure is likely to occur, and current and historical data regarding emission sources.
√ √
tobacco smoke should be removed as completely as possible.
√ X
historical exposures in a range of settings are needed to improve the characterization of uncertainties, both quantitative and qualitative, in historical models of exposures.
√ √
Design needs for new studies of exposure- response
with respect to magnitude, frequency, and duration, rather than solely by duration of employment.
regulatory concern, including a range of exposures to provide a base for understanding the relation between exposure and health effects.
√ √ √ √
quantified where possible;
account in both power calculations and exposure response analyses.
√ √ √- √-
factors for this disease.
a case-control or case-cohort design
√ √ X X
21
22
23
Daniel Krewski, Chair HEI Panel University of Ottawa
the evaluation criteria and research needs identified in 1999 by HEI
underground workers)
2
their value for risk assessment
3
documented and scientifically justified to test the study hypotheses…
Case series/ Case reports Cross-sectional studies Case-Control studies Cohort studies RCTs
its specific design features and conduct.
Increasing strength of design Truckers DEMS 4
5
the Panel concluded that there was no evidence indicating significant confounding that would invalidate the application of the Truckers or DEMS studies for quantitative risk assessment
would be the ability to understand the implications of any gaps (i.e., smoking and radon) on the degree of uncertainty in the exposure-response relationship
6
underground and were likely to be most highly exposed?
7
P-value for trend .006 .062
Odds ratios and 95% confidence intervals for cumulative REC lagged 15 years, by mining facility without adjustment for radon (Silverman 2012, Table 7)
8
for complete cohort
9
T-1 Trend P-value 0.02 0.02 0.03 0.03 0.02 0.03
Cumulative Exposure Lagged 15 Years, adjusted for radon “with radon” models, for all subjects after omitting data from a single mine (Crump et al 2015, Table VI). Same test for trend as Silverman et al. 2012
Mine A
>547.5 25.4 to <547.5 2.0 to <25.4 0 to <2.0
Mine B
>547.6 76.2 to <547.6 3.4 to <76.2 0 to <3.4
Mine D
>460.6 49.7 to <460.6 1.6 to <49.7 0 to <1.6
Mine E
>508.6 56.8 to <508.6 3.4 to <56.8 0 to <3.4
Mine G
>579.2 82.9 to <579.2 3.4 to <82.9 0 to <3.4
Mine H
>563.4 92.5 to <563.4 6.8 to <92.5 0 to <6.8 OR (95% CI) 1 2 3 4 5 6 7 8 9 10
Mine I
>535.7 87.5 to <535.7 3.4 to <87.5 0 to <3.4
Mine Omitted
Limestone Potash Potash Salt Salt Trona Trona
0.06 Mine Type Does the Type of Mine Matter? Analyses by Crump et al 2015 10
Tests for Trend p-value T1: 0.02 T2: 0.72 T1: 0.05 T2: NT* T1: 0.29 T2: 0.95
Odds Ratios and Trend Tests Based on Cumulative Exposure Lagged 15 years, Adjusted for Radon, by work location (Crump et al 2015). T1 test- average exposure; T2 – Continuous exposure assignments *NT: Negative trend (not significant)
Analyses by Crump et al 2015
11
DEMS
that any one mine could explain all effects (although one analysis of the cohort data suggested that the mine with the highest level of exposure [Limestone mine A] had somewhat higher risks)
12
has accrued since the previous panel reported on this issue in 1999 is both relevant and informative.
13
individually and collectively provide useful new information that advances our understanding of the relationship between the exposure to diesel exhaust experienced by the workers in those studies and their risk of lung cancer.
14
Risk characterization: modeling, assumptions, adjustments uncertainties:
demographics
personal exposures over a lifetime
levels and composition Image of general population What is the observed risk of lung cancer? What is the predicted risk of lung cancer? Other relevant data methods, and analyses 15
16
treatment of all data sets
exposure metrics (REC, SEC, EC)
Vermeulen et al 2013 17
the composition and toxicity of diesel exhaust
Advanced Collaborative Emissions (ACES Project) 18
contributions to EC component of PM2.5 (1986-2015)
19 Diesel PM PM2.5 EC
µg/m3
Los Angeles Basin
sufficiently robust to develop quantitative assessments of human lung cancer risks, and
estimate risks at lower concentrations than observed in
20
21
HEAL TH EFFECTS INSTITUTE: DIRECTOR OF SCIENCE – RASHID SHAIKH PROJECT MANAGER – KATY WALKER PRINCIP AL SCIENTIST –AARON COHEN REVIEW SCIENTIST – KATE ADAMS RESEARCH ASSIST ANT – ADAM CERVENKA PUBLISHING – CAROL MOYER UNIVERSITY OF OTT AWA: NAGARAJ YENUGADHATI YUANLI SHI
22